FYI, more fun with floats: > 0.1+0.1==0.2 [1] TRUE > 0.1+0.1+0.1+0.1==0.4 [1] TRUE > 0.1+0.1+0.1==0.3 [1] FALSE > 0.1+0.1+0.1==0.1*3 [1] TRUE > 0.3==0.1*3 [1] FALSE
¯\_(ツ)_/¯ But this is not R's fault. See: https://0.30000000000000004.com Iñaki El vie., 31 ago. 2018 a las 15:36, Iñaki Ucar (<iu...@fedoraproject.org>) escribió: > > El vie., 31 ago. 2018 a las 15:10, Felix Ernst > (<felix.gm.er...@outlook.com>) escribió: > > > > Dear all, > > > > I a bit unsure, whether this qualifies as a bug, but it is definitly a > > strange behaviour. That why I wanted to discuss it. > > > > With the following function, I want to test for evenly space numbers, > > starting from anywhere. > > > > .is_continous_evenly_spaced <- function(n){ > > if(length(n) < 2) return(FALSE) > > n <- n[order(n)] > > n <- n - min(n) > > step <- n[2] - n[1] > > test <- seq(from = min(n), to = max(n), by = step) > > if(length(n) == length(test) && > > all(n == test)){ > > return(TRUE) > > } > > return(FALSE) > > } > > > > > .is_continous_evenly_spaced(c(1,2,3,4)) > > [1] TRUE > > > .is_continous_evenly_spaced(c(1,3,4,5)) > > [1] FALSE > > > .is_continous_evenly_spaced(c(1,1.1,1.2,1.3)) > > [1] FALSE > > > > I expect the result for 1 and 2, but not for 3. Upon Investigation it turns > > out, that n == test is TRUE for every pair, but not for the pair of 0.2. > > > > The types reported are always double, however n[2] == 0.1 reports FALSE as > > well. > > > > The whole problem is solved by switching from all(n == test) to > > all(as.character(n) == as.character(test)). However that is weird, isn’t it? > > > > Does this work as intended? Thanks for any help, advise and suggestions in > > advance. > > I guess this has something to do with how the sequence is built and > the inherent error of floating point arithmetic. In fact, if you > return test minus n, you'll get: > > [1] 0.000000e+00 0.000000e+00 2.220446e-16 0.000000e+00 > > and the error gets bigger when you continue the sequence; e.g., this > is for c(1, 1.1, 1.2, 1.3, 1.4, 1.5, 1.6, 1.7): > > [1] 0.000000e+00 0.000000e+00 2.220446e-16 2.220446e-16 4.440892e-16 > [6] 4.440892e-16 4.440892e-16 0.000000e+00 > > So, independently of this is considered a bug or not, instead of > > length(n) == length(test) && all(n == test) > > I would use the following condition: > > isTRUE(all.equal(n, test)) > > Iñaki > > > > > Best regards, > > Felix > > > > > > [[alternative HTML version deleted]] > > > > ______________________________________________ > > R-devel@r-project.org mailing list > > https://stat.ethz.ch/mailman/listinfo/r-devel > > > > -- > Iñaki Ucar -- Iñaki Ucar ______________________________________________ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
